At 3:35 AM -0700 7/24/97, James B. Byrne wrote:
>We have a couple of what are for us very busy lists that have a
>large number of AOL subscribers. On a regular basis AOL seems
>to stop accepting e-mail traffic for periods of up to several
>hours and on one recent occasion more than a day. This
>apparently drives our mail sever resource consumption way up,
>although I am not clear as to why.
Then is sounds like your mail system isn't configured well, and I would
fix that, not try to hack around it with TLB.
What do you maen by consumption going way up? Unfortunately, it's not
at all clear what you mean by this. Is the loadaverage going up? Too
many sendmails running and forcing the system into paging? what?
I'm guessing, but... chances are, from the sound of it, you think there
are too many sendmails running and it's affecting system performance.
>We are running Sendmail 8.8.5 MJ 1.94.1 and TLB.
First step would be to force sendmail NOT to constantly retry queue
items. Sendmail can be configged to leave messages alone for a while
between tries, so it's not constantly beating its head on a down server:
# minimum time in queue before retry
O MinQueueAge=30m
At the least you should have this set to 30m, just on general
principles. (I run my sites with a very short -bd -q5, and minqueueage
to 30m. sendmail walks the queue often, but doesn't waste time on
messages tried in the previous half hour. Very little loading problem
here....). There's no reason not to up this to 1 hour to cut the load
even more if you want.
Next, check your loadaverages. If they're growing because of sendmail
daemons, simply drop your queueing limits:
# load average at which we just queue messages
#O QueueLA=8
(since my load average rarely goes above 1, this is obviously not a
problem for my box, but...)
You can also experiment with these:
# load average at which we refuse connections
#O RefuseLA=12
# maximum number of children we allow at one time
#O MaxDaemonChildren=12
# maximum number of new connections per second
#O ConnectionRateThrottle=3
But the latter two work better mostly for keeping incoming connections
under control, not mail-list generated loads. For that, I prefer the
above stuff.
No offense, but I think you need to understand what the problem is
before you fix it. It sort of sounds here like you've got a flat tire,
and you're trying to fix it by giving the car a tuneup... If you're
problem is queue processing, AOL isn't the cause, it's a symptom. You
can hack around with with AOL -- but it'll come back the next time a
large site starts burping...
--
Chuq Von Rospach (chuq@apple.com) Apple IS&T Mail List Gnome
<http://www.solutions.apple.com/>
Plaidworks Consulting (chuqui@plaidworks.com) <http://www.plaidworks.com/>
(<http://www.plaidworks.com/hockey/> +-+ The home for Hockey on the net)
Follow-Ups:
References:
|
|
